Upload
hoangdien
View
221
Download
0
Embed Size (px)
Citation preview
Artificial Intelligence for Networked Robotic Drones
C.A. Rabbath
Precision Weapons SectionDefence Research and Development Canada Valcartier
Jan 31 2011
1
Semi-Autonomy or, in certain cases, complete autonomy will be a feature of all future platforms and systems
Future robotics will become increasingly modular, reconfigurable, self-repairing, and self-sustaining
Micro air vehicles will be used as short-range weapons and ISR collectors, and will able to use swarming and linking tactics
Some projections from the CF- Projecting Power: Canada’s Air Force 2035, CFAWC.
2
Artificial Intelligence
… the study and design of intelligent agents [1]
[1] Russell, Stuart J.; Norvig, Peter (2003), Artificial Intelligence: A Modern Approach (2nd ed.), Upper Saddle River, New Jersey: Prentice Hall
… an intelligent agent is a system that perceives its environmentand takes actions that maximize its chances of success [1]
Here – AI for a collective of robotic drones
… coordination of the motion of several UAVs
… AI enables UAV group to take actions with limited humanintervention
3
• Persistent Surveillance (24/7)• ISTAR• Search and Rescue• Air Lift• Air Refueling• Coordinated bombing• Protection of an area• …Air liftAir lift
SurveillanceSurveillance
SARSAR
Air refuelingAir refueling
Networked Robotic Drones
Some Envisaged Applications (CFAWC, USAF documents)
4
AI – Drone formation flight
AISoftware
• * Formation maintains prescribed geometries in flight
* Formation efficiently handles degraded conditions and
unexpected events on its own
* Formation complies to high-level commander’s commands
without continuous human supervision
* Our AI design allows for scalability
The commander DOES NOT pilot the UAVs
AIS adjusts to damage, motor problems, adverse weather, crash of a drone,
drone takeover …
In-line
V
From 2-vehicle squadron to 100-vehicle SWARM
5
Bayesian FilteringAlternative
communicationsConsensus techniques
Hypothesis TestingObserver DesignPrognosticsDiagnosis
Collective Artificial IntelligenceTo act
To obtain information (Sensing / Networked)
To detect abnormal conditions(AI software)
Decentralized OptimizationDynamic ProgrammingHeuristicsLearningStochastic Games
Faults/failures in actuators, control surface, sensors
Body damage Adverse environmental effects& Changes in environment
To decide(AI software)
6
Collective Artificial Intelligence – Drone formation faced with an anomaly
LeaderFollower 1
Follower 2
Monitoring (Sensors + AI software)Follower 1 goes into abnormal motion
Follower 2 detects anomaly and adapts its motion/positioning (AI software)
7
Collective Artificial Intelligence –Drone formation shape change
V-Shape request by GCSTeam adaptation (AI software)
8
Collective Artificial Intelligence –Robust formation of drones
Adaptation by the team (AI software)Perturbation to leader
9
10
EXTRA SLIDES
11
Collective Artificial Intelligence –Mobile Sensor Network
* Dynamic Reconfiguration* Dynamic Reconfiguration
* Sensor Placement Optimization* Sensor Placement Optimization
EventsX